All RCC, CC, and EG were raised in ponds at the State Key Laboratory of Developmental Biology of Freshwater Fish, Hunan Normal University, Changsha, China. NG were purchased at a local market. All experiments were approved by the Animal Care Committee of Hunan Normal University and followed the guidelines of the Administration of Affairs Concerning Animal Experimentation of China.

To compare the sequences of the chordin gene between single-tail and twin-tail fish, we isolated and sequenced the 1^st to 6^th exons of chordin homologues from the embryonic cDNA pools of four fish: single-tail RCC and CC, and twin-tail EG and NG. After self-mating, total RNA was extracted from the gastrula-stage embryos of all four fish using Trizol (Invitrogen, Camarillo, CA, United States). The first-strand cDNA of the chordin gene was synthesized using ReverTra Ace (Toyobo, Osaka, Japan). The forward primer 5′-GCGTTACCCATCCAACC-3′ and the reverse primer 5′-TCTGTRTCCGCTTGTGGT-3′ were designed based on CDS of the chordin genes from Carassius auratus (AB874473.1) and Cyprinus carpio (LC092194.1), which were downloaded from GenBank. Each PCR (25 μL) contained 20 ng of cDNA template, 1.5 mM MgCl₂, 0.2 mM of each dNTP, 0.4 μM of each primer, 1 × PCR buffer, and 1.25 U Taq polymerase (Takara, Dalian, China). The cycling conditions were as follows: an initial denaturation at 94°C for 4 min; followed by 30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 1 min; and a final extension at 72°C for 10 min. PCR products were separated and purified using 1.2% agarose gels and Gel Extraction Kits (Sangon, Shanghai, China), respectively. Purified products were ligated into pMD18-T vectors and transfected into Escherichia coli DH5α. Positive clones were sequenced and further analyzed using BLAST¹ and CLUSTAL W².

We randomly selected eight fish of each type (RCC, CC, NG, and EG) for SSR testing. Total genomic DNA was extracted from the fin tissue of each fish with a DNA extraction kit (Sangon, Shanghai, China). The microsatellite regions were PCR amplified using six florescent-labeled microsatellite primers (synthesized by Sangon, Shanghai, China): three from the common carp (MWF 4, MWF 5, and MWF 16) (Crooijmans et al., 1997); two from the bighead carp (HLJY 3940 and HLJY 2526) (Geng et al., 2006); and one from the crucian carp (MFW1, developed for this study). Primer sequences and annealing temperatures are listed in Supplementary Table 1. PCR amplifications were performed in a total volume of 20 μL, containing 1 μL of genomic DNA (5 ng/μL), 2 μL of 10 × Taq Buffer (Mg²⁺ Plus), 1 μL of 2.5 mM dNTPs, 0.4 μL of each primer (5 μM), 0.4 μL of Taq DNA Polymerase (5 U/μL), and 14.8 μL of ddH₂O. The PCR cycling conditions were as follows: an initial denaturation at 94°C for 3 min; followed by 35 cycles of 94°C for 30 s, a primer-specific annealing temperature for 30 s, and 72°C for 45 s; a final extension step of 72°C for 7 min. PCR products were sequenced using capillary electrophoresis on an ABI 3730XL DNA sequencer (Applied Biosystems, Foster City, CA, United States) using BigDye Terminator Cycle Sequencing kits (Applied Biosystems, Foster City, CA, United States).

Genetic distance and genetic polymorphism indexes, including major allele frequency, numbers of genotypes, numbers of alleles, heterozygosity, and gene diversity, were calculated using Popgene32 (Yeh et al., 2000). SSR genotypes, polymorphic information content (PIC) values and the cluster analysis of the 4 populations, which was performed by UPGMA method basing on the genetic distance, were determined with PowerMarker 3.25 (Liu and Muse, 2005).

Both RCC and CC have single-tail (Figures 1A,B), while EG and NG have twin-tail (Figures 1C,D). To investigate the formation of the twin-tail phenotype in EG, the 1^st to 6^th exons of chordin homologues were cloned from RCC, CC, EG, and NG. All cloned sequences were identified as chordin and were submitted to GenBank (RCC, MN918649; CC, MN918650; EG, MN918651; NG, MN918652). There was high nucleotide sequence similarity (> 99%) among the chordin genes from all four fish (Supplementary Figure 1). Further analysis showed that chordin gene sequences from RCC and CC, the single-tail fish, were identical, as were the chordin sequences of EG and NG, the twin-tail fish. However, the chordin sequences between the twin-tail and single-tail fish differed: the twin-tail fish had a stop codon (TAG) at the 127^th amino acid position of the chordin protein, while the single-tail fish had a glutamic acid codon (GAG) (Figure 2).

Across the four fish (EG, NG, RCC, and CC), each of the six amplified SSR loci (121–302 bp) had 1–8 alleles (Table 1). Almost all the alleles identified in EG, as well as most in NG, were also found in RCC (Table 1). However, nearly all the alleles (except one in HLJY 2526) identified in CC were absent in EG and RCC. The peaks at the HLJY 2526 loci were identical across all four fish (Figure 3A). However, at the MWF 16 locus, RCC, EG, and NG had a peak at 275 bp, while CC had a peak at 133 bp instead (Figure 3B). The peak patterns at the MFW 1 loci presented similar situation as MWF 16 where CC was different from RCC, NG and EG (Figure 3F), and higher similarity to HLJY3940 (Figure 3E). Like RCC and NG, EG had a peak at 206 bp at allele HLJY 3940; however, unlike RCC and NG, EG lacked a peak at 241 bp (Figure 3E). Indeed, all EG alleles also appeared in RCC. In contrast, NG exhibited a specific allele of MWF 5 (at 185 bp) (Figure 3D). In addition, NG respectively presented one allele similar to CC at MWF 4 (at 164 bp) (Figure 3C) and MWF 5 (at 156 bp) (Figure 3D). These results indicated that, although EG and NG appeared morphologically similar, these fish differed genetically. Genetic polymorphism analyses indicated that, of the four populations investigated (RCC, CC, NG, and EG), EG had the lowest polymorphism indexes, corresponding to the highest homogeneity (Table 2). In addition, across all pairs of taxa, genetic distance was lowest between EG and RCC (0.1103; Table 3). Consistent with this, the UPGMA phylogenetic tree recovered EG and RCC as a sister taxa. NG, EG, and RCC formed a single cluster, distinct from CC (Figure 4).

Hybridization is an important method of animal breeding because this process increases genetic variation (Bullini, 1994), and because hybrids frequently present transgressive segregation exceeding the range between the parental means (Slatkin and Lande, 1994; Rieseberg et al., 1999). The morphological characteristics of EG were highly similar to those of NG, but differed from the original parents RCC and CC: unlike RCC and CC, EG had twin-tail, a spherical body, a short caudal peduncle, and a range of body coloration patterns (Wang et al., 2014). Of these morphological characters, the most recognizable is the twin-tail, as most other fish species have a single tail.

In NG, the twin-tail trait is associated with mutations in the 4^th exon of the chordin gene (Takashima et al., 2007; Abe et al., 2014). In NG, two alleles of the chordin A gene have been identified: the wild-type allele (chordin A^wt) includes a glutamic codon at the 127^th amino-acid position, while the other allele (chordinA^E127X) includes a stop codon at the same position. The allele chordinA^E127X is predicted to encode a truncated protein that contributes to the formation of the twin-tail (Abe et al., 2014). In zebrafish, dysfunction in the chordin gene gave rise to a twin-tail phenotype (Schulte-Merker et al., 1997; Fisher and Halpern, 1999). In EG, the potential primary generation (G₀) of NG, we identified the twin-tail-associated mutation at the 127^th amino-acid position in the chordin gene: a stop codon (TAG) instead of a glutamic acid codon (GAG). This mutation was identical to that found in NG. In contrast, the chordin genes of RCC and CC had a wild-type glutamic acid codon at the same position (Figure 2). These results in the G₀ generation of EG are consistent with previous studies of twin-tail formation (Abe et al., 2014), and provide direct evidence that this mutation, as well as the associated phenotype (i.e., twin-tail), might have arisen due to hybridization.

SSRs (or microsatellite markers) have been broadly used in studies of molecular evolution (Chistiakov et al., 2006). These simple tandem-repeats are abundantly and randomly distributed throughout genomes, and due to their high polymorphism and rapid evolutionary rate, microsatellite loci are useful for analyses of genetic diversity, population structure, gene flow, and hybridization (Pritchard et al., 2000; Wang et al., 2014). SSR genotyping, which is based on the numbers of repeats, may improve our understanding of genetic relationships within populations or among close relatives (Roy et al., 1994). SSR allele inheritance proceeds in accordance with Mendelian laws of segregation and independent assortment. That is, in diploids, each offspring inherits one copy of each microsatellite segment from each parent (Tesson et al., 2013). Thus, EG, derived from the hybridization of RCC and CC, should have inherited both sets of parental microsatellites. However, among the six alleles, which showed good reproducibility and stability between RCC and CC, we only identified RCC alleles in EG; all alleles unique to CC were completely lost (Table 1 and Figure 3). It may be that the inheritance of SSR alleles by EG did not conform to Mendelian laws because the EG lineage was subject to intense artificial selection. As a result of this selective pressure, the EG population did not contain all the alleles from RCC and CC.

Large-scale morphological changes, which require extensive modifications of developmental mechanisms, are often presumed to require relatively long periods of evolutionary time (Janvier, 2002; Kardong, 2011). However, previous studies have shown that gene duplication and subsequent artificial selection generated dramatic morphological and developmental changes in NG strains within 600 years (Abe et al., 2014). Previously, using a combination of hybridization and artificial selection, we established a stable EG lineage in even less time (F₁–F₆, within 6 years) (Wang et al., 2014). Here, our results suggested that the hybridization of female RCC and male CC led to a mutation in the chordin gene, which contributed to the expression of the twin-tail phenotype. During hybridization, the separation, recombination, and mutation of genetic materials leads to the formation of new varieties (Schumer et al., 2016). Thus, our results reveal the effects of hybridization on the formation of the twin-tail phenotype in EG, and SSR inheritance patterns, shedding new light on the origin and evolution of NG.